DocVQA leaderboard
DocVQA
1 models tested · Updated 2026-03-11 · Verified sources only
Qianfan-OCR leads at 92.8%
1
Qianfan-OCR
Baidu · arxiv/2603.13398 · 2026-03-11
Unified end-to-end model. Strong DocVQA performance from OCR-specialized architecture.
92.8%
benchmark.space canonical public ledger of AI capability